Batch Process Modeling with Multilayer Recurrent Fuzzy Neural Network

摘要

A multilayer recurrent fuzzy neural network (MRFNN) with local feedbacks is proposed for batch process modeling. The local feedbacks in the membership layer and the rule layer introduce dynamics into the network.Learning algorithm of MRFNN include structure learning and parameters learning. By structure learning, the membership and rule layers are automatically constructed. Modified chaotic search (CS) and least square estimation (LSE) are combined for parameters learning,where CS is for tuning the premise parameters including feedback coefficients of the membership and rule layers, and LSE is for updating the consequent coefficients accordingly. Results of simulation on nonlinear function identification and a batch reactor reveal that the proposed MRFNN can capture the nonlinear and time-varying characteristics of dynamic system well.